Вы находитесь на странице: 1из 9

ARTICLE IN PRESS

Energy Policy 36 (2008) 3077 3085

Contents lists available at ScienceDirect

Energy Policy
journal homepage: www.elsevier.com/locate/enpol

A re-examination of the relationship between electricity consumption and


economic growth in Malaysia
Chor Foon Tang 
Media Utara Resources, 675, Mk4, Bukit Merah, Permatang Pauh, 13500 Penang, Malaysia

a r t i c l e in fo

abstract

Article history:
Received 26 September 2007
Accepted 25 April 2008
Available online 6 June 2008

The purpose of this study is to re-investigate the relationship between electricity consumption and
economic growth in Malaysia from 1972:1 to 2003:4. This study adopted the newly developed ECMbased F-test [Kanioura, A., Turner, P., 2005. Critical values for an F-test for cointegration in the
multivariate model. Applied Economics 37(3), 265270] for cointegration to examine the presence of
long run equilibrium relationship through the autoregressive distributed lag (ARDL) model. The
empirical evidence suggests that electricity consumption and economic growth are not cointegrated in
Malaysia. However, the standard Grangers test and MWALD test suggest that electricity consumption
and economic growth in Malaysia Granger causes each other. This nding provides policymakers with a
better understanding of electricity consumption and allows them to formulate electricity consumption
policy to support the economic development and to enhance the productivity of capital, labour and
other factors of production for future economic growth in Malaysia.
& 2008 Elsevier Ltd. All rights reserved.

Keywords:
ECM
Electricity consumption
Growth

1. Introduction
The issue of economic prosperities have always been linked to
the international trade and the inow of foreign direct investment. However, these were the age-old issues in the economic
literature. Over the past decades, one of the most debated topic
abroad is the relationship between electricity consumption and
economic growth. Ferguson et al. (2000) found that there is a
strong correlation between electricity consumption and economic
growth. This is consistent with Keyness economic framework
where consumption and income are signicantly correlated. In the
earlier studies, cointegration and Granger causality tests have
been extensively used to examine the presence of long run
equilibrium and the direction of causality between electricity
consumption and economic growth. Unfortunately, the existing
empirical studies failed to provide a consensus causality results.
Some empirical studies claimed that economic growth induces
electricity consumption to change (Chen et al., 2007), while others
defended the view that electricity consumption leads to economic
growth through the Keynesian multiplier effect (Shiu and Lam,
2004). Jumbe (2004) noted that the causal link between
electricity consumption and economic growth is of concern
because it has a signicant policy implication in a countrys
economic development. For instance, if electricity consumption

 Tel.: +60 12 423 2254; fax: +60 4 390 1533.

E-mail address: tcfoon@yahoo.com


0301-4215/$ - see front matter & 2008 Elsevier Ltd. All rights reserved.
doi:10.1016/j.enpol.2008.04.026

Granger causes economic growth, the goal of development policy


should encourage electricity consumption. However, if economic
growth is not the result of electricity consumption, policy
initiatives that encourage electricity consumption could be
detrimental to the economy. Therefore, it is of paramount
importance for this study to afrm the causal relationship
between electricity consumption and economic growth. The
major contribution of this study is to apply the newly developed
cointegration and Granger causality tests to re-investigate the
existence of long run equilibrium relationship and causality
direction between electricity consumption and economic growth
in Malaysia. The electricity consumption in Malaysia is of interest
because of the following: rst, Malaysias electricity consumption
is the second highest among the ve ASEAN founding economies.1
The electricity consumption per capita grows rapidly since 1971
and this may be one of the important factors leading Malaysias
economic growth. The plots of the electricity consumption per
capita in kilowatt hours (kWh) and the real Gross National
Product (GNP) are depicted in Fig. 1.
From the visual inspection, we observed that both series are
moving smoothly with an upward trend over the period of
19722003. During the period of 19902000, electricity consumption per capita in Malaysia has increased from 1194.26 kWh

1
ASEAN refers to Association of Southeast Asian Nations and the ve ASEAN
founding economies are Indonesia, Malaysia, the Philippines, Singapore and
Thailand.

ARTICLE IN PRESS
3078

C.F. Tang / Energy Policy 36 (2008) 30773085

9
Log of Electricity Consumption per capita
Log of Real Income per capita
8

4
1975

1980

1985

1990

1995

2000

Fig. 1. The plots of log of electricity consumption (kWh) per capita and real gross
national products per capita.

per capita to 2774.07 kWh per capita. This may be due to the rapid
development of information and communication technologies
(ICTs), and other infrastructures that consume high power
electricity in Malaysia. Thus, electricity is a vital input of power
for Malaysias economic growth and development, and her vision
to become a developed country in year 2020.
Second, to the best of our knowledge, the study on electricity
consumption in Malaysia is relatively few, and limited to Yoo
(2006) and Chen et al. (2007). Both studies consistently suggested
that the variables are not cointegrated using the conventional
tests (Engle and Granger, 1987; Johansen and Juselius, 1990).
Therefore, these studies employed Granger causality test based on
rst difference vector autoregressive (VAR) model to determine
the direction of causality between electricity consumption and
economic growth in Malaysia. In this respect, those existing
studies only provide information on short run causality.
Third, Johansen cointegration technique was widely used in
examining the long run equilibrium relationship between electricity consumption and economic growth. Nevertheless, couple
studies (e.g. Reimers, 1992; Cheung and Lai, 1993) have conducted
Monte Carlo experiment to examine the nite sample performance of Johansen cointegration test. These studies found that in
nite samples Johansens cointegration test is bias toward
rejecting the null hypothesis of no cointegration. Besides that,
Huang and Yang (1996) noted that the Johansen procedure is very
sensitive to the assumption that the errors are independent
normal. When the errors are not independent normal, the
Johansen test has a tendency to nd spurious cointegration.
Moreover, the Gonzalo and Lees (1998) simulation results
indicate that Johansens likelihood ratio (LR) test tend to nd
spurious cointegration with probability approaching to one when
the order of integration of the estimated series are not purely I(1)
process. Abeysinghe and Tan (1999) demonstrated that Johansen
estimator was the worst among six cointegrating estimators (e.g.
Engle and Granger, 1987; Bardsen, 1989; Johansen and Juselius,
1990; Phillips and Hansen, 1990; Engle and Yoo, 1991; Stock and
Watson, 1993). In addition, they noted that Johansen estimator is
highly sensitive to the lag length of the VAR and the deterministic
components (i.e. constant and trend) included into the cointegrating equations. In line to that, Ahking (2002) and Hjelm and
Johansson (2005) showed that the results of Johansen tests are
highly sensitive to the choice of deterministic components. In
practice, Johansen (1992) recommended Pantula procedure to

select a proper model for cointegration test, however, Hjelm and


Johansson (2005) argued that this procedure cannot overcome the
problem effectively as the procedure tend to select Model 3 (i.e. an
unrestricted constant). Therefore, the Johansens cointegration
test results provided by the earlier studies may be biased due to
the aforementioned shortcomings.
In this study, we attempt to re-investigate the relationship
between electricity consumption and economic growth in
Malaysia to ll the detected gaps mentioned above. This study
differs from the existing studies in at least three dimensions: First,
we employ the Gandolfo (1981) annual to quarterly data
interpolation technique to overcome the nite sample size
problem. Zhous (2001) Monte Carlo evidence shows that
increasing the data frequency will improve the statistical test
properties. Thus, the use of high-frequency data through the
interpolation technique may improve the power of the statistical
test used in this study.
Second, to ascertain the long run equilibrium relationship
between electricity consumption and economic growth in
Malaysia we employ the newly developed error-correction model
(ECM)-based F-test procedure (Kanioura and Turner, 2005), within
an autoregressive distributed lag (ARDL) framework. If the
variables are cointegrated, Bardsens (1989) method will be used
to derive the short and long run coefcients from the estimated
ARDL model. Third, the earlier studies on electricity consumption
and economic growth in Malaysia only provide the short run
causality with the VAR model because the variables are not
cointegrated. In this study, we will perform the modied Wald
(MWALD) test through the augmented VAR model (Toda and
Yamamoto, 1995) to determine the long run causality direction
between electricity consumption and economic growth in
Malaysia. One of the most important features of this test is that
without knowing the information of the order of integration or
cointegration, the test remains valid in examining the long run
causality with data at the level (see Bhattacharya and Mukherjee,
2002a, b; Lau and Chan, 2003). Furthermore, the use of MWALD
test may reduce the likelihood of making wrong decision on the
orders of integration and cointegration. Another advantage of
using MWALD test is because of its simplicity. On the other hand,
the Grangers test with VAR model will be used to determine the
direction of short run causality.
The remainder of this paper is organised in the following
manner. A brief literature reviews will be presented in Section 2.
The discussion on the data and econometric techniques used in
this study are demonstrated in Section 3. Lastly, Sections 4 and 5
report the empirical results and conclusion, respectively.

2. Literature reviews
The relationship between energy consumption and economic
growth was rst discussed by Kraft and Kraft (1978) with the
United States data from 1947 to 1974. They found that the
causality runs from economic growth to energy consumption.
Over the past decades, researchers have extended their idea to
examine the relationship between electricity consumption and
economic growth. The earlier empirical ndings of the causality
direction between electricity consumption and economic growth
can be summarised in Table 1. A conclusion that we can drawn
from Table 1 is that the causal link between electricity consumption and economic growth remains controversial.
Altinay and Karagol (2005) employed the MWALD test to
examine the causality direction between electricity consumption
and economic growth in Turkey over the period of 19502000.
The MWALD test evidence showed that electricity consumption
Granger causes economic growth in Turkey, but there is no

Table 1
The summaries of the cointegration and causality results of the earlier studies
No.

Authors

Countries

Study period

Econometric techniques
Cointegration

Empirical results
Causality

Cointegration

Causality

Turkey

19502000

Toda and Yamamoto (1995)MWALD

EC-EG

Pakistan

19551996

Engle-Granger (1987)

Hsiaos (1981)VAR

Not cointegrated

EC-EG

3.

Chen et al. (2007)

China
Hong Kong
India
Indonesia
Korea
Malaysia
Philippines
Singapore
Taiwan
Thailand

19712001

Johansen-Juselius (1990)

Granger (1988)VECM

Cointegrated
Cointegrated
Cointegrated
Cointegrated
Cointegrated
Not cointegrated
Cointegrated
Cointegrated
Cointegrated
Cointegrated

ECVEG
EC-EG
EG-EC
ECVEG
ECVEG
EG-EC
EG-EC
EG-EC
ECVEG
ECVEG

4.

Ghosh (2002)

India

19501997

Johansen-Juselius (1990)

Granger (1969)VAR

Not cointegrated

EG-EC

5.

Jumbe (2004)

Malawi

19701999

Engle-Granger (1987)

Granger (1988)VECM

Cointegrated

EC2EG

6.

Morimoto and Hope (2004)

Sri Lanka

19601998

Granger (1969)VAR

EC2EG

7.

Mozumder and Marathe (2007)

Bangladesh

19711999

Johansen-Juselius (1990)

Granger (1988)VECM

Cointegrated

EG-EC

Shiu and Lam (2004)

Shanghai, China

19712000

Johansen-Juselius (1990)

Granger (1988)VECM

Cointegrated

ECVEG

Squalli (2006)

Algeria
Indonesia
Iran
Iraq
Kuwait
Libya
Nigeria
Qatar
Saudi Arabia
UAE
Venezuela

19802003

Pesaran et al. (2001)Bounds test

Toda and Yamamoto (1995)MWALD

Cointegrated
Cointegrated
Cointegrated
Cointegrated
Not cointegrated
Cointegrated
Cointegrated
Cointegrated
Cointegrated
Cointegrated
Not cointegrated

EG-EC
EC-EG
EC-EG
EG-EC
EG-EC
EG-EC
EC-EG
EC-EG
EC2EG
EC-EG
EC-EG

10.

Wolde-Rufael (2006)

Algeria
Benin
Cameroon

19712001

Pesaran et al. (2001)Bounds test

Toda and Yamamoto (1995)MWALD

Not cointegrated
Cointegrated
Cointegrated

ECVEG
EC-EG
EG-EC

3079

8.
9.

ARTICLE IN PRESS

Altinay and Karagol (2005)


Aqeel and Butt (2001)

C.F. Tang / Energy Policy 36 (2008) 30773085

1.
2.

3080

Table 1 (continued )
No.

Authors

Countries

Study period

Econometric techniques
Cointegration

Empirical results
Causality

Congo, DR.
Congo, Rep.
Egypt
Gabon
Ghana
Kenya
Morocco
Nigeria
Senegal
South Africa
Sudan
Tunisia
Zambia
Zimbabwe
Yang (2000)

Taiwan

19541997

Engle-Granger (1987)

Granger (1969)VAR

Causality

Not cointegrated
Cointegrated
Not cointegrated
Cointegrated
Not cointegrated
Not cointegrated
Cointegrated
Cointegrated
Not cointegrated
Cointegrated
Not cointegrated
Not cointegrated
Cointegrated
Cointegrated

EC-EG
ECVEG
EC-EG
EC-EG
EG-EC
ECVEG
EC-EG
EG-EC
EG-EC
ECVEG
ECVEG
EC-EG
EG-EC
EG-EC

Not cointegrated

EC2EG

Yuan et al. (2007)

China

19782004

Johansen-Juselius (1990)

Granger (1988)VECM

Cointegrated

EC-EG

Yoo (2005)

Korea

19702002

Johansen-Juselius (1990)

Granger (1988)VECM

Cointegrated

EC-EG

14.

Yoo (2006)

Indonesia
Malaysia
Singapore
Thailand

19712002

Engle-Granger (1987) and Johansen-Juselius (1990)

Granger (1969)VAR and Hsiaos (1981)VAR

Not
Not
Not
Not

EG-EC
EC2EG
EC2EG
EG-EC

cointegrated
cointegrated
cointegrated
cointegrated

Note: EC and EG denotes electricity consumption and economic growth, respectively. - represents the unidirectional causality. The 2 and V imply bilateral causality, and no causality, respectively.

ARTICLE IN PRESS

12.
13.

C.F. Tang / Energy Policy 36 (2008) 30773085

11.

Cointegration

ARTICLE IN PRESS
C.F. Tang / Energy Policy 36 (2008) 30773085

evidence on reverse causal link. This implied that electricity


consumption played an important role in fostering Turkeys
economic development. Squalli (2006) conducted a study on
electricity consumption and economic growth in Organization of
Petroleum Exporting Countries (OPEC) members through the
cointegration and causality analyses. This study used annual data
from 1980 to 2003. The author suggests that electricity consumption and economic growth are cointegrated in all countries
through the evidence of bounds testing procedure to cointegration
developed by Pesaran et al. (2001). Furthermore, the MWALD test
results exhibited mixed causality evidence among OPEC members.
The causality evidence suggests that electricity consumption
Granger causes economic growth in the case of Indonesia, Nigeria,
UAE and Venezuela. On the other hand, economic growth is found
to Granger causes electricity consumption in the case of Algeria,
Iraq, Kuwait and Libya. Ultimately, the bilateral causality evidence
is only found in the case of Qatar and the Saudi Arabia.
Using the similar cointegration and causality approaches,
Wolde-Rufael (2006) conducted a study on 17 African countries
with annual data from 1971 to 2001. The cointegration results
between electricity consumption and economic growth are nonconsensus among African countries. Specically, only 50 per cent
of the selected African countries are cointegrated (see Table 1).
Similarly, the MWALD test evidences are also non-consensus
among the selected countries. The author found that electricity
consumption and economic growth are not related for Algeria,
Congo Republic, Kenya, South Africa and Sudan. This is contrary to
Squalli (2006) ndings in the case of Algeria. In an empirical
survey, Engel (1996) found that different sample period and/or
statistical method employed may yield different statistical
inference or nding. Therefore, one plausible reason for this
conicting result may be due to the different sample size
employed as they used the same causality test.
Yoo (2006) employed the Engle and Granger (1987) and the
Johansen and Juselius (1990) cointegration tests to investigate the
presence of long run relationship between electricity consumption and economic growth in four ASEAN countries, hereafter
ASEAN-4. Both cointegration tests results consistently suggest
that electricity consumption and economic growth are not
cointegrated for ASEAN-4 countries over the period of
19712002. The Hsiaos version of Granger causality test results
indicated that there is bilateral causality between electricity
consumption and economic growth for Malaysia and Singapore.
On the other hand, economic growth Granger causes electricity
consumption for Indonesia and Thailand, but there is no reverse
causality evidence for these two countries. A more recent paper,
Chen et al. (2007) employed the Johansens test to examine the
presence of long run equilibrium relationship between electricity
consumption and economic growth in Asian countries over the
period of 19712001. The Johansens test results suggest that the
variables are cointegrated for all the countries, except Malaysia. As
the variables are not cointegrated for Malaysia, they employed the
Grangers causality test based on rst difference VAR model to
investigate the short run causality direction between the interested variables. The estimation results suggest that the short run
causality is running from income to electricity consumption for
Malaysia, but there is no evidence for reverse causality. This
causality evidence contradicted with Yoos (2006) nding, which
is more reliable as the causality direction should be bilateral. This
is because high electricity consumption is require to support the
expansionary of industrialisation, transportation and manufacturing sectors development, which eventually lead to Malaysias
economic growth and vice versa. In view of conicting causality
result for Malaysia, difference in sample period and causality
testing procedure may be the plausible explanations (see Engel,
1996).

3081

3. Data and econometric techniques


3.1. Data
This study uses quarterly data of Malaysias electricity
consumption per capita in kWh and real GNP per capita from
1972:1 to 2003:4 (T 128). The quarterly data were interpolated
using Gandolfo (1981) technique because the data for electricity
consumption and GNP for Malaysia are available only in annual
basis. This interpolation technique is chosen because it has been
widely accepted and used in many of the published empirical
studies. Among them are Habibullah (1998), Baharumshah and
Rashid (1999) and Baharumshah et al. (2006). Furthermore, Smith
(1998) conducted Monte Carlo analysis to evaluate the effects of
the linearly interpolating technique on cointegration tests and he
found that the interpolated series does not cause any bias in the
estimates of cointegrating vectors even when the sample size is
small. The Gandolfos interpolation procedure is cited in Appendix
A. The annual raw data for interpolate was extracted from World
Bank, World Development Indicators (WDI). The Consumer Price
Index (CPI, 2000 100) is used to deate the nominal GNP per
capita into the real term.
3.2. Unit root test
The standard augmented DickeyFuller (ADF) and PhillipsPerron (PP) unit root tests have been criticised for its low power
in distinguishing between unit root and a near unit root process
(Campbell and Perron, 1991; DeJong et al., 1992). The low power of
ADF and PP unit root tests has prompted this paper to use the
Kwiatkowski et al. (1992)KPSS semi-parametric unit root test to
conrm the degree of integration. The KPSS semi-parametric
procedure tests for the null hypothesis of level (Zm) or trend (Zt)
stationarity against the alternative hypothesis of non-stationarity.
The advantage of using KPSS test is that it has superior properties
in small sample (Kwiatkowski et al., 1992). Furthermore, the test
is able to distinguish between unit root and a near unit root
process. As a result, KPSS test is perform better than ADF and PP
unit root tests. The following expressed the KPSS testing equation:
LM Zm Zt

T
X

s2 kT 2

t1

S2t

(1)

P
where St ti1 u^ i ; u^ t are the estimated residuals from the
ordinary least squares (OLS) from a regression of the time series
on a constant and a linear deterministic trend. If no linear trend is
present in the time series is only regressed on a constant. s2(k) is
the non-parametric estimate of the long run variance of u^ t and k
represents for the lag truncation parameter.
3.3. Kanioura and Turners (2005) ECM-based F-test for
cointegration
The ECM-based F-test for cointegration procedure has been
proposed by Kanioura and Turner (2005). This procedure poses
better statistical properties over the residual-based two-step
EngleGranger approach because it does not push the short run
dynamic into the residual terms. Kanioura and Turners (2005)
Monte Carlo experiment evidence shows that the ECM-based Ftest is more powerful than the two-step EngleGranger cointegration test. Moreover, Cooks (2006) Monte Carlo evidence also
supports that the ECM-based F-test has greater power than the
EngleGranger and the generalised least squares (GLS) cointegration tests. Therefore, the ECM-based F-test is considered more
reliable than the conventional cointegration tests. In order to
apply this test the order of integration must be uniform. Thus,

ARTICLE IN PRESS
3082

C.F. Tang / Energy Policy 36 (2008) 30773085

pre-testing of unit root is necessary to determine the order of


integration. To implement the ECM-based F-test for cointegration
we estimate the ARDL model as presented in Eq. (2).
DLEC t a f TRENDt p1 LEC t1 p2 LEGt1
p
p
X
X

di DLEC ti
lj DLEGtj t
i1

(2)

j0

where D is the rst difference operator and p is the maximum lag


order. The TREND variable is included into the testing model
because both of the series showed a strong trending pattern (see
Fig. 1). The LEC and LEG represents the natural logarithm of
electricity consumption per capita and real gross national product
per capita, respectively. The existence of long run equilibrium
relationship is tested by restricting the lagged level variables,
LECt1 and LEGt1 in the Eq. (2). It is a joint signicance F-test for
the null hypothesis of no cointegrating relation (H0: p1 p2 0)
against the alternative hypothesis of a cointegrating relation (H1:
p16p260). If the computed F-statistics is greater than the critical
values tabulated in Kanioura and Turner (2005, p. 267) we reject
the null hypothesis of no cointegrating relation. Otherwise,
electricity consumption and economic growth are not cointegrated.
If the variables are cointegrated, Bardsens (1989) method will
be used to compute the short and long run elasticities. Bardsen
(1989) noted that this approach is computational convenient and
provides efcient result. The long run coefcient is derived as the
coefcient of the one lagged level explanatory variable divided by
the coefcient of the one lagged level dependent variable and then
multiplies with a negative sign. Thus, the long run coefcient for
real income is (p2/p1). On the other hand, the short run effect is
captured by the total coefcients of the rst differenced variables
P
in Eq. (2). Thus, the short run coefcient for real income is
lj .
3.4. Causality tests
In this study, we employed the Granger and MWALD causality
tests to gauge the short and long run causality through the
following VAR and augmented VAR models, respectively. This
study will not offer a detail explanation on Grangers test because
it has been well dened in the existing literature. However, this
study do not include the deterministic trend variable into the VAR
model for testing the short run causality as there is usually no
trending pattern occur in rst differenced series.2 To implement
the short run causality through the Grangers causality test, we
estimate the Eq. (3) and the F-test is use for restriction on the
parameters of the VAR model. From Eq. (3), the alternative
hypothesis, A12,k608k implies that economic growth Granger
causes electricity consumption; whereas if A21,k608k holds, the
electricity consumption Granger causes economic growth in the
short run.
The VAR model for Granger testshort run causality:
"
# " # "
# "
#
A11;1 A12;1
a1
DLEC t1
DLEC t


A21;1 A22;1
a2
DLEGt1
DLEGt
"
# "
#
A11;k A12;k
DLEC tk


A21;k A22;k
DLEGtk
"
#
x1t

(3)
x2t
For the long run causality, we use the MWALD test developed by
Toda and Yamamoto (1995). Zapata and Rambaldi (1997)
2

We thank to the anonymous referees for the suggestion.

documented that both likelihood ratio test and Wald test are
very sensitive to the specication of the short run dynamics in
error-correction model (ECM) even in large samples. Furthermore,
they noted that given the performance of the tests in larger
samples, the MWALD test has better appeal because of its
simplicity. In order to employ the MWALD test, we pre-specify
the maximal order of integration (dmax) for the series in the
system and the optimal lags order (k) for the VAR. We use
dmax 1 because it performs better than other orders of dmax (see
Dolado and Lutkepohl, 1996). To ascertain the causality direction
between electricity consumption and economic growth, we
estimate the following augmented VAR(k+dmax) models shown
in Eq. (4). The optimal lag length k is determined by Schwarz
Bayesian Criterion (SBC).
The augmented VAR model for MWALD testlong run
causality:
"
# " # "
#
"
#
B11;1 B12;1
j1
a1
LEC t

 TRENDt 
B21;1 A22;1
j2
a2
LEGt
"
#
"
#
B11;k B12;k
LEC t1


B21;k B22;k
LEGt1
"
# "
#
B11;p B12;p
LEC tk


B21;p B22;p
LEGtk
"
# "
#
LEC tp
x1t


(4)
LEGtp
x2t
where D is the rst difference operator, k is the optimal lag order
and p represents the (k+dmax) lag orders. From Eq. (4), B12,k608k
implies that the economic growth Granger causes electricity
consumption; whereas if B21,k608k holds, the electricity consumption Granger causes economic growth in the long run.

4. Empirical results
The unit root tests results are summarised in Table 2.
At 5 per cent signicance level, the result of KPSS test indicates
that all the estimated variables are non-stationary at level, but it is
stationary after rst differencing. This implies that real income
per capita (LEG) and electricity consumption per capita (LEC) are
integrated of order one, I(1) process. These results are consistent
with the notion that most of the macroeconomic variables are
non-stationary at level, but become stationary after rst differencing (Nelson and Plosser, 1982).
With this nding we can proceed with the ECM-based F-test
for cointegration. A common practice in performing ECM-based Ftest for cointegration is to determine the optimal lag order. The
SBC statistic indicates that ARDL(9, 5) is the best lag orders
combination and the estimation outputs are reported in Table 3.
Batteries of diagnostic tests were performed on the nal ARDL
model. In particular, the Ramsey RESET test failed to reject the null
hypothesis of no general specication error. Thus, the model is
free from specication error. The BreuschGodfrey LM test
exhibited that the residuals are not serially correlated. Unfortunately, the plot of CUSUM of Squares statistic crossed the 5 per
cent critical bounds (see Fig. 2). Thus, the estimated coefcients
are not stable over the period of 19861995. The rapid expansionary of export-orientated and manufacturing industries in
Malaysia since 1980s may be the plausible explanation of this
structural break. This is also in line with Malaysian government
policy to attract more inux of foreign direct investment into
high-technology manufacturing sectors. Thus, the electricity
consumption in Malaysia may grow rapidly and hence caused
this structural break.

ARTICLE IN PRESS
C.F. Tang / Energy Policy 36 (2008) 30773085

3083

Table 2
The results of unit root tests
Variables

ADF

PP

KPSS

Tm

tt

Z(tm)

Z(tt)

Zm

Zt

Level
LECt
LEGt

1.247 (9)
0.902 (9)

3.124 (9)
2.770 (9)

1.198 (7)
1.206 (5)

1.178 (7)
2.453 (5)

1.371 (9)*
1.334 (9)*

0.116 (9)
0.080 (9)

First difference
DLECt
DLEGt

1.327 (8)
2.873 (8)***

1.480 (8)
2.876 (8)

4.522 (9)*
5.244 (9)*

4.610 (9)*
5.233 (9)*

0.197 (7)
0.088 (5)

0.139 (7)***
0.053 (5)

Note: The t, Z(t) and Z statistics refer to the ADF, PP and KPSS unit root tests, respectively. The subscript m and t indicate the models that allow for intercept term and both an
intercept and a deterministic trend, respectively. The asterisks *, ** and *** denotes the signicance level at 1, 5 and 10 per cent, respectively. The optimal lag length for ADF
test is selected using SBC while the bandwidth for PP and KPSS tests are selected using Newey-West Bartlett kernel. Figures in parentheses indicate the optimal lag length
and bandwidth. The critical values for ADF and PP tests are obtained from MacKinnon (1996) while the asymptotic critical values for KPSS test are obtained from
Kwiatkowski et al. (1992).

Table 3
The results of error-correction model-based F-test
Variables

Lag length for ARDL

ECM-based F-test

LEC, LEG, TREND

(9, 5)

4.336

Critical values
1 per cent
5 per cent
10 per cent

8.12
5.94
4.86

Note: The optimal lag length is selected by SBC. The asterisks *, ** and *** denotes
the signicance level at 1, 5 and 10 per cents, respectively. The critical values in
above are obtained from Kanioura and Turner (2005, p. 267) Table 1.
R2: 0.851; adjusted R2 0.824.
F-statistics: 31.511 (0.00).
Jarque-Bera: 7.701 (0.02).
Breusch-Godfrey LM test: 4.376 (0.112).
Ramsey RESET test: 0.681 (0.409).
( ) refer to p-value.

In order to ascertain the presence of long run equilibrium


relationship between electricity consumption and economic
growth, a joint signicance F-test for H0: p1 p2 0 was
performed. The computed F-statistic is 4.336 that smaller than
the 10 per cent critical value, indicating that electricity consumption and economic growth in Malaysia are not cointegrated. Thus,
these variables are not coalescing in the long run, and this is
consistent with Yoo (2006), and Chen et al. (2007) cointegration
test results. A remarkable nding from this study is that
increasing the data frequency and the application of more advance
cointegration technique may not alter the cointegration results.
Next, the Grangers and MWALD tests are used in this study to
determine the direction of short and long run causality even
though the variables are not cointegrated.3 These causality tests
3
Traditionally, if the economic variables are not cointegrated, then one can
determine the short run causality based on rst difference VAR model. In contrast,
ECM will be specied to determine the short run causality (signicant of the rst
differenced series) and long run causality (signicant of the lagged level errorcorrection term) if the variables are cointegrated. These denitions imply that the
rst differenced series is used to measure the short run causal effect, while the
series at level is used to measure the long run causal effect. The MWALD test
developed by Toda and Yamamoto (1995) and Dolado and Lutkepohl (1996) has
been extensively used to investigate the long run causality as the causal inference
was conducted in the level augmented-VARs that may contain integrated and
(non-) cointegrated processes (e.g. Bhattacharya and Mukherjee, 2002a, b; Lau and
Chan, 2003). In order words, the MWALD test can be used to examine the long run
causality irrespective of whether the variables are cointegrated or not as the causal
relationship was conducted in the level augmented-VARs. Therefore, it is perhaps
not surprising that long run causality can exist even though the variables are not
cointegrated because the cointegration and causality tests serve different purposes

are sensitive to the lag structure incorporated with the VAR


model. Therefore, we employed the SBC to choose the appropriate
lag structure of the VAR model. The results of the causality
analyses are reported in Table 4. For the short run causality, the
Grangers test results showed that the electricity consumption
and economic growth in Malaysia Granger causes each other.
These causality ndings signicantly support the Yoo (2006)
results, but contrary to Chen et al. (2007), which found only one
direction of causality. A new contribution from this study is the
evidence of a set of long run causality through the MWALD testing
procedure. Consistently, the MWALD test results indicate that
electricity consumption and economic growth in Malaysia showed
bilateral causality. This nding exhibit that without knowing the
order of integration and cointegration of the variables we surmise
that the electricity consumption and economic growth Granger
causes each others in the long run. Interestingly, both causality
tests consistently imply that Malaysia is an energy-dependent
country (i.e., LEC-LEG), thus efcient use of energy and energy
development policies will have positive effect on economic
growth. This nding is in line to the rapid development of ICTs,
manufacturing, and other infrastructures industries that consume
high power electricity in Malaysia. Thus, electricity is an
important input of power for Malaysias economic growth and
development toward the vision of becoming a developed country
in the year 2020. In this respect, better management on energy
growth policies are require to ensure that the electricity is
sufcient to support the Malaysias economic development.

5. Conclusions
The intention of this study is to re-investigate the role of
electricity consumption on economic growth in Malaysia over the
period of 1972:12003:4. Due to the nite sample size problem,
this study adopted the Gandolfo (1981) interpolation technique to
derive the quarterly data from the existing annual raw data. An
interesting nding from this study is that the increased data
frequency did not improve the cointegration result. This is in
line with Hakkio and Rush (1991) ndings that conceptually
(footnote continued)
(see Dufour and Renault, 1998; Dufour et al., 2006, for detail explanation of long
run causality). For example Wolde-Rufael (2006) found that electricity consumption and economic growth are not cointegrated, but MWALD test shows evidence
of either unidirectional or bilateral causal relationship between electricity
consumption and economic growth in the case of Egypt, Ghana, Senegal and
Tunisia. In addition to that, the author also shows that the presence of
cointegration does not necessarily implied causation (e.g. in the case of Congo
Republic and South Africa).

ARTICLE IN PRESS
3084

C.F. Tang / Energy Policy 36 (2008) 30773085

1.2

30
CUSUM

5% Significance

1.0

20

CUSUM of Squares

5% Significance

0.8

10

0.6
0
0.4
-10

0.2

-20

0.0

-30

-0.2
80 82 84 86 88 90 92 94 96 98 00 02

80 82 84 86 88 90 92 94 96 98 00 02

Fig. 2. Plots of CUSUM and CUSUM of squares tests.

Appendix A. Quarterly interpolation from annual


observations

Table 4
The results of Granger and MWALD tests
Null hypotheses

Grangers test (short run)

MWALD test (long run)

w -statistics
LEGQLEC
LECQLEG

22.672*
14.981*

31.151*
21.086*

Note: The asterisks *, ** and *** denotes signicance level at 1, 5 and 10 per cent,
respectively. Q represents does not Granger causes.

cointegration is a long run phenomenon and by merely increasing


the data frequency, it will not alter the cointegration result.
Therefore, the ndings of this study suggest that electricity
consumption and economic growth are not moving together in
the long run.
However, the standard Grangers and MWALD tests reported
bilateral causality evidences between electricity consumption and
economic growth in Malaysia. These imply that electricity
consumption is an important element for Malaysias economic
development as Malaysia is an energy-dependent country. This
high demand of electricity is parallel with Malaysias economic
policy toward the vision of becoming an industrialised and
developed country in the year 2020. Therefore, sufcient supply
of electricity is required to support the industrial development
and to enhance the productivity of capital, labour and other
factors of production.
In view of policy implication, the ndings of this study suggest
that electricity consumption played an important role in economic
development. Therefore, policies to manage the supply of
electricity are required to ensure that the electricity is sufcient
to support Malaysias economic development. In addition to that,
the Ministry of Water, Energy and Communications department
should continue to explore new resources and expand the
electricity supply via hydroelectricity power station in Malaysia
because it is environmental friendly, renewable, cost-effective and
stable compared with fossil fuel.

Acknowledgements
The author would like to thank the anonymous referees and
Nicky France (Editor) of the Energy Policy for their insightful
comments and suggestions. The usual disclaimer applies.

Since electricity consumption and Gross National Products


(GNP) are not available in quarterly series, the Gandolfo (1981)
interpolation method was used in this study to compute the
quarterly series from annual data. In undertaking the derivation of
the interpolation formula, the observed values are actually
integrals. Therefore, we must integrate the quadratic function in
order to obtain the quarterly formula. After satisfying all the
conditions in any year t, the formula to compute the quarterly
series are presented as follows:
1st Quarter : y1
t 0:0546875yt1 0:234375yt
 0:0390625yt1

(A.1)

2nd Quarter : y2
t 0:0078125yt1 0:265625yt
 0:0234375yt1

(A.2)

3rd Quarter : y3
t  0:0234375yt1 0:265625yt
0:0078125yt1

(A.3)

4th Quarter : y4
t  0:0390625yt1 0:234375yt
0:0546875yt1

(A.4)

where yt, yt1 and yt+1 are the current, lag and lead values of the
annual data. For instance, to obtain the rst quarter series for
electricity consumption, one may substitute the annual electricity
consumption values for yt, yt1 and yt+1 into Eq. (A.1). The similar
way was used to compute the series for second, third and fourth
quarters, respectively. It is worth to point out here that, two
annual observations will be loss due to the one period lead and lag
values of the annual data. Therefore, we are able to compute the
quarterly series from 1972:1 to 2003:4 if the annual data from
1971 to 2004 are used to interpolate.
References
Abeysinghe, T., Tan, K.B., 1999. Small sample estimation of a cointegrating vector:
an empirical evaluation of six estimation techniques. Applied Economics
Letters 6, 645648.
Ahking, F.W., 2002. Model mis-specication and Johansens co-integration
analysis: an application to the US money demand. Journal of Macroeconomics
24, 5166.
Altinay, G., Karagol, E., 2005. Electricity consumption and economic growth:
evidence from Turkey. Energy Economics 27 (6), 849856.
Aqeel, A., Butt, M.S., 2001. The relationship between energy consumption and
economic growth in Pakistan. Asia-Pacic Development Journal 8 (2), 101109.

ARTICLE IN PRESS
C.F. Tang / Energy Policy 36 (2008) 30773085

Baharumshah, A.Z., Rashid, S., 1999. Export, imports and economic growth in
Malaysia: empirical evidence based on multivariate time series. Asian
Economic Journal 13 (4), 389406.
Baharumshah, A.Z., Lau, E., Khalid, A.M., 2006. Testing twin decits hypothesis
using VARs and variance decomposition. Journal of the Asia Pacic Economy 11
(3), 331354.
Bardsen, G., 1989. Estimation of long-run coefcients in error-correction models.
Oxford Bulletin of Economics and Statistics 51 (3), 345350.
Bhattacharya, B., Mukherjee, J., 2002a. Nature of causal relationship between stock
market and macroeconomic aggregates in Indiaan empirical analysis. In:
Chakravarty, C. (Ed.), Financial Markets: Perspectives and Approach. Indian
Institute of Capital Markets, Mumbai.
Bhattacharya, B., Mukherjee, J., 2002b. Causal relationship between stock market
and exchange rate, foreign exchange reserves and value of trade balance: a case
study for India. Available at /http://www.igidr.ac.in/money/basabi.pdfS.
Campbell, J.Y., Perron, P., 1991. Pitfalls and opportunities: what macroeconomists
should know about unit roots. NBER Macroeconomic Annual 6, 141201.
Chen, S.T., Kou, H.I., Chen, C.C., 2007. The relationship between GDP and electricity
consumption in 10 Asian countries. Energy Policy 35 (4), 26112621.
Cheung, Y.W., Lai, K.S., 1993. Finite-sample sizes of Johansens likelihood ratio tests
for cointegration. Oxford Bulletin of Economics and Statistics 55 (3), 313328.
Cook, S., 2006. The power of single equation tests for cointegration. Applied
Economics Letters 13 (5), 265267.
DeJong, D.A., Nankervis, J.C., Savin, N.E., Whiteman, C.H., 1992. Integration versus
trend stationarity in time series. Econometrica 60 (2), 423433.
Dolado, J.J., Lutkepohl, H., 1996. Making Wald tests work for cointegrated VAR
system. Econometric Reviews 15 (4), 369386.
Dufour, J.M., Renault, E., 1998. Short run and long run causality in time series:
theory. Econometrica 66, 10991125.
Dufour, J.M., Pelletier, D., Renault, E., 2006. Short run and long run causality in time
series: inference. Journal of Econometrics 132, 337362.
Engel, C., 1996. The forward discount anomaly and the risk premium: a survey of
recent evidence. Journal of Empirical Finance 3, 123192.
Engle, R.F., Granger, C.W.J., 1987. Cointegration and error-correction: representation, estimation and testing. Econometrica 55 (2), 251276.
Engle, R.F., Yoo, B.S., 1991. Cointegrated economic time series: an overview with
new results. In: Engle, R.F., Granger, C.W.J. (Eds.), Long run Economic
Relationships. Oxford University Press, New York, pp. 237266.
Ferguson, R., Wilkinson, W., Hill, R., 2000. Electricity use and economic
development. Energy Policy 28 (13), 923934.
Gandolfo, G., 1981. Quantitative Analysis and Econometric Estimation of Continuous Time Dynamic. North-Holland, Amsterdam.
Ghosh, S., 2002. Electricity consumption and economic growth in India. Energy
Policy 30 (2), 125129.
Gonzalo, J., Lee, T.H., 1998. Pitfalls in testing for long run relationships. Journal of
Econometrics 86 (1), 129154.
Granger, C.W.J., 1969. Investigating causal relations by econometric models and
cross-spectral methods. Econometrica 37 (3), 424438.
Granger, C.W.J., 1988. Some recent development in the concept of causality. Journal
of Econometrics 39 (12), 199211.
Habibullah, M.S., 1998. Money, output and stock prices in Malaysia: further
evidence. Borneo Review 9 (2), 135155.
Hakkio, C.S., Rush, M., 1991. Cointegration: how short is the long-run? Journal of
International Money and Finance 10 (4), 571581.
Hjelm, G., Johansson, M.W., 2005. A Monte Carlo study on the pitfalls in
determining deterministic components in cointegrating models. Journal of
Macroeconomics 27, 691703.
Hsiao, C., 1981. Autoregressive modeling and money-income causality detection.
Journal of Monetary Economics 7 (1), 85106.

3085

Huang, B.N., Yang, C.W., 1996. Long run purchasing power parity revisited: a Monte
Carlo simulation. Applied Economics 28 (8), 967974.
Johansen, S., 1992. Determination of cointegration rank in the presence of a linear
trend. Oxford Bulletin of Economics and Statistics 54, 383397.
Johansen, S., Juselius, K., 1990. Maximum likelihood estimation and inference on
cointegration with applications to the demand for money. Oxford Bulletin of
Economics and Statistics 52 (2), 169210.
Jumbe, C.B.L., 2004. Cointegration and causality between electricity consumption
and GDP: empirical evidence from Malawi. Energy Economics 26 (1), 6168.
Kanioura, A., Turner, P., 2005. Critical values for an F-test for cointegration in the
multivariate model. Applied Economics 37 (3), 265270.
Kraft, J., Kraft, A., 1978. On the relationship between energy and GNP. Journal of
Energy Development 3 (2), 401403.
Kwiatkowski, D., Phillips, P.C.B., Schmidt, P., Shin, Y., 1992. Testing the null of
stationarity against the alternative of a unit root: how sure are we that the
economic time series have a unit root? Journal of Econometrics 54 (13),
159178.
Lau, E., Chan, T.H., 2003. Transmission mechanism of twin decits hypothesis:
evidence from neighbouring countries. INTI Journal 1 (3), 159166.
MacKinnon, J.G., 1996. Numerical distribution functions for unit root and
cointegration tests. Journal of Applied Econometrics 11 (6), 601618.
Morimoto, K., Hope, C., 2004. Impact of electricity supply on economic growth in
Sri Lanka. Energy Economics 26 (1), 7785.
Mozumder, P., Marathe, A., 2007. Causality relationship between electricity
consumption and GDP in Bangladesh. Energy Policy 35 (1), 395402.
Nelson, C.R., Plosser, C.I., 1982. Trends and random walks in macroeconomic time
series: some evidence and implications. Journal of Monetary Economics 10 (2),
139162.
Pesaran, M.H., Shin, Y., Smith, R.J., 2001. Bounds testing approaches to the analysis
of level relationships. Journal of Applied Econometrics 16 (3), 289326.
Phillips, P.C.B., Hansen, B.E., 1990. Statistical inference in instrumental variables
regression with I(1) processes. Review of Economic Studies 57, 99125.
Reimers, H.E., 1992. Comparison of tests for multivariate cointegration. Statistical
Papers 33, 335359.
Shiu, A., Lam, P.L., 2004. Electricity consumption and economic growth in China.
Energy Policy 32 (1), 4754.
Smith, S.F., 1998. Cointegration tests when data are linearly interpolated.
Unpublished Paper, State University of New York at Albany.
Squalli, J., 2006. Electricity consumption and economic growth: bounds and
causality analyses of OPEC members. Energy Economics 29 (6), 11921205.
Stock, J.H., Watson, M., 1993. A simple estimator of cointegrating vectors in higher
order of integrated system. Econometrica 61, 783820.
Toda, H.Y., Yamamoto, T., 1995. Statistical inference in vector autoregressions with
possibly integrated process. Journal of Econometrics 66 (12), 225250.
Wolde-Rufael, Y., 2006. Electricity consumption and economic growth: a time
series experience for 17 African countries. Energy Policy 34 (10), 11061114.
Yang, H.Y., 2000. A note of the causal relationship between energy and GDP in
Taiwan. Energy Economics 22 (3), 309317.
Yoo, S.H., 2005. Electricity consumption and economic growth: evidence from
Korea. Energy Policy 33 (12), 16271632.
Yoo, S.H., 2006. The causal relationship between electricity consumption and
economic growth in the ASEAN countries. Energy Policy 34 (18), 35733582.
Yuan, J., Zhao, C., Yu, S., Hu, Z., 2007. Electricity consumption and economic growth
in China: cointegration and co-feature analysis. Energy Economics 29 (6),
11791191.
Zapata, H.O., Rambaldi, A.N., 1997. Monte Carlo evidence on cointegration and
causation. Oxford Bulletin of Economics and Statistics 59 (2), 285298.
Zhou, S., 2001. The power of cointegration tests versus data frequency and time
spans. Southern Economic Journal 67 (4), 906921.